Apple Microarchitecture Research by Dougall Johnson

M1/A14 P-core (Firestorm): Overview | Base Instructions | SIMD and FP Instructions
M1/A14 E-core (Icestorm):  Overview | Base Instructions | SIMD and FP Instructions

RSHRN2 (4S)

Test 1: uops

Code:

  rshrn2 v0.8h, v1.4s, #3
  movi v0.16b, 1
  movi v1.16b, 2

(no loop instructions)

1000 unrolls and 1 iteration

Retires: 1.000

Issues: 1.000

Integer unit issues: 0.001

Load/store unit issues: 0.000

SIMD/FP unit issues: 1.000

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)dispatch simd uop (57)ldst uops in schedulers (5b)dispatch uop (78)map simd uop (7e)map simd uop inputs (81)? int output thing (e9)? simd retires (ee)
1004303310011100010007590510001000200011000
1004303310011100010007590510001000200011000
1004303310011100010007590510001000200011000
1004303310011100010007590510001000200011000
1004303310011100010007590510001000200011000
1004303310011100010007590510001000200011000
1004303310011100010007590510001000200011000
1004303310011100010007590510001000200011000
1004303310011100010007590510001000200011000
1004303310011100010007590510001000200011000

Test 2: Latency 1->1

Code:

  rshrn2 v0.8h, v1.4s, #3
  movi v0.16b, 1
  movi v1.16b, 2

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code): 3.0033

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)dispatch int uop (56)dispatch simd uop (57)int uops in schedulers (59)ldst uops in schedulers (5b)dispatch uop (78)map int uop (7c)map simd uop (7e)map int uop inputs (7f)map simd uop inputs (81)? int output thing (e9)? simd retires (ee)? int retires (ef)
1020430033101011011000010010000300768905101002001000420020012110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100

1000 unrolls and 10 iterations

Result (median cycles for code): 3.0033

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)dispatch int uop (56)dispatch simd uop (57)int uops in schedulers (59)ldst uops in schedulers (5b)dispatch uop (78)map int uop (7c)map simd uop (7e)map int uop inputs (7f)map simd uop inputs (81)? int output thing (e9)? simd retires (ee)? int retires (ef)
10024300331002121100002010000707689051002020100062020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707692471005120100442020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010

Test 3: Latency 1->2

Code:

  rshrn2 v0.8h, v0.4s, #3
  movi v0.16b, 1

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code): 3.0033

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)dispatch int uop (56)dispatch simd uop (57)int uops in schedulers (59)ldst uops in schedulers (5b)dispatch uop (78)map int uop (7c)map simd uop (7e)map int uop inputs (7f)map simd uop inputs (81)? int output thing (e9)? simd retires (ee)? int retires (ef)
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100
1020430033101011011000010010000300768905101002001000420020008110000100

1000 unrolls and 10 iterations

Result (median cycles for code): 3.0033

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)dispatch int uop (56)dispatch simd uop (57)int uops in schedulers (59)ldst uops in schedulers (5b)dispatch uop (78)map int uop (7c)map simd uop (7e)map int uop inputs (7f)map simd uop inputs (81)? int output thing (e9)? simd retires (ee)? int retires (ef)
10024300331002121100002010000707689051002020100042020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010
10024300331002121100002010000707689051002020100002020000111000010

Test 4: throughput

Count: 8

Code:

  movi v0.16b, 0
  rshrn2 v0.8h, v8.4s, #3
  movi v1.16b, 0
  rshrn2 v1.8h, v8.4s, #3
  movi v2.16b, 0
  rshrn2 v2.8h, v8.4s, #3
  movi v3.16b, 0
  rshrn2 v3.8h, v8.4s, #3
  movi v4.16b, 0
  rshrn2 v4.8h, v8.4s, #3
  movi v5.16b, 0
  rshrn2 v5.8h, v8.4s, #3
  movi v6.16b, 0
  rshrn2 v6.8h, v8.4s, #3
  movi v7.16b, 0
  rshrn2 v7.8h, v8.4s, #3
  movi v8.16b, 9

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code divided by count): 0.5011

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)dispatch int uop (56)dispatch simd uop (57)int uops in schedulers (59)ldst uops in schedulers (5b)dispatch uop (78)map int uop (7c)map simd uop (7e)map int uop inputs (7f)map simd uop inputs (81)? int output thing (e9)? simd retires (ee)? int retires (ef)
1602044048280109101800081008001230032005280112200800122001600241160000100
1602044011080110101800091008001330032005280112200800122001600241160000100
1602044008680109101800081008001230032005280112200800122001600241160000100
1602044008680109101800081008001230032005280112200800122001600241160000100
1602044009980110101800091008001330032005280112200800122001600241160000100
1602044008680109101800081008001230032005280112200800122001600241160000100
1602044008680109101800081008001230032005280112200800122001600241160000100
1602044008680109101800081008001230032005280112200800122001600241160000100
1602044008680109101800081008001230032005280112200800122001600241160000100
1602044008680109101800081008001230032005280112200800122001600241160000100

1000 unrolls and 10 iterations

Result (median cycles for code divided by count): 0.5054

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)dispatch int uop (56)dispatch simd uop (57)int uops in schedulers (59)ldst uops in schedulers (5b)dispatch uop (78)map int uop (7c)map simd uop (7e)map int uop inputs (7f)map simd uop inputs (81)? int output thing (e9)? simd retires (ee)? int retires (ef)
1600244379180020118000910800133032005280022208001220160000116000010
1600244191280011118000010800003032000080010208000020160000116000010
1600244043280011118000010800003032000080010208000020160000116000010
1600244039380011118000010800003032000080010208000020160000116000010
1600244041080011118000010800003032000080010208000020160000116000010
1600244039380011118000010800003032000080010208000020160000116000010
1600244041780011118000010800003032000080010208000020160000116000010
1600244041880011118000010800003032000080010208000020160000116000010
1600244042180011118000010800003032000080010208000020160000116000010
1600254417980057118004610800503032005280022208001220160094116000010

Test 5: throughput

Count: 16

Code:

  rshrn2 v0.8h, v16.4s, #3
  rshrn2 v1.8h, v16.4s, #3
  rshrn2 v2.8h, v16.4s, #3
  rshrn2 v3.8h, v16.4s, #3
  rshrn2 v4.8h, v16.4s, #3
  rshrn2 v5.8h, v16.4s, #3
  rshrn2 v6.8h, v16.4s, #3
  rshrn2 v7.8h, v16.4s, #3
  rshrn2 v8.8h, v16.4s, #3
  rshrn2 v9.8h, v16.4s, #3
  rshrn2 v10.8h, v16.4s, #3
  rshrn2 v11.8h, v16.4s, #3
  rshrn2 v12.8h, v16.4s, #3
  rshrn2 v13.8h, v16.4s, #3
  rshrn2 v14.8h, v16.4s, #3
  rshrn2 v15.8h, v16.4s, #3
  movi v16.16b, 17

(fused SUBS/B.cc loop)

100 unrolls and 100 iterations

Result (median cycles for code divided by count): 0.5002

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)dispatch int uop (56)dispatch simd uop (57)int uops in schedulers (59)ldst uops in schedulers (5b)dispatch uop (78)map int uop (7c)map simd uop (7e)map int uop inputs (7f)map ldst uop inputs (80)map simd uop inputs (81)? int output thing (e9)? ldst retires (ed)? simd retires (ee)? int retires (ef)
16020480086160105101160004100160008300640036160108200160012200032002810160000100
16020480335160303101160202100160206300640036160108200160012200032002410160000100
16020480035160105101160004100160008300640036160108200160012200032002410160000100
16020580070160143101160042100160054300640036160108200160012200032002410160000100
16020480331160304101160203100160207300640036160108200160012200032002410160000100
16020480035160105101160004100160008300640036160108200160012200032002410160000100
16020480035160105101160004100160008300640036160108200160012200032002410160000100
16020480035160105101160004100160008300640036160108200160012200032002410160000100
16020480035160105101160004100160008300640036160108200160012200032002410160000100
16020480035160105101160004100160008300640036160108200160012200032002410160000100

1000 unrolls and 10 iterations

Result (median cycles for code divided by count): 0.5002

retire uop (01)cycle (02)schedule uop (52)schedule int uop (53)schedule simd uop (54)dispatch int uop (56)dispatch simd uop (57)int uops in schedulers (59)ldst uops in schedulers (5b)dispatch uop (78)map int uop (7c)map simd uop (7e)map int uop inputs (7f)map simd uop inputs (81)? int output thing (e9)? simd retires (ee)? int retires (ef)
160024801901600151116000410160008306400361600182016001220320000116000010
160024800391600111116000010160000306400001600102016000020320124116000010
160024800351600111116000010160000306400001600102016000020320000116000010
160024800351600111116000010160000306400001600102016000020320000116000010
160024800351600111116000010160000306400001600102016000020320000116000010
160024800351600111116000010160000306400001600102016000020320000116000010
160024800351600111116000010160000306400001600102016000020320000116000010
160024800351600111116000010160000306400001600102016000020320000116000010
160024800351600111116000010160000306400001600102016000020320130116000010
160024800351600111116000010160000306400001600102016000020320000116000010